Last week news broke that a young Google engineer—a Romeoville native and University of Illinois graduate—had written a memo that was going viral within the company. A criticism of the company's diversity initiatives, it leaned heavily on evolutionary psychology to argue that women, on average, aren't as suited to programming and other high-stress and high-status positions as men, which he claimed explains the gender imbalances found there and elsewhere within the tech field.

In the midst of the debate his memo set off, he was quickly fired, generating a second debate: Did the memo create a hostile working environment by implicitly suggesting that the author's colleagues are unqualified?

Social psychology has shown plenty of evidence to that effect: By generalizing that women are bad at math, you make individual women worse at actually doing math than those who aren't exposed to that stereotype.

It's a phenomenon called stereotype threat, and it's been detected for other social groups as well. In 2007, University of Chicago psychologist Sian Beilock—recently named president of Barnard College—and two co-authors designed a test to examine exactly how it works. The first part was a pretty simple examination of stereotype threat: They gave female college students with good math skills a test, told some of them they were participating in a study about why women aren't as good at math as men, and didn't tell the others that. Those who were reminded of the stereotype scored about 80 percent, compared with 90 percent among the control group—a drop of a full letter grade.

Beilock and her co-authors added a twist to figure out how, specifically, stereotypes hinder performance. First, they presented math problems in ways that required different levels of verbal processing. Then, they tested subjects on verbal and spacial problems.

Participants exposed to stereotypes had a tougher time working on the math problems presented verbally, and also underperformed on verbal problems—suggesting, researchers said, that the stereotype threat hinders performance by suppressing someone's verbal ability. It also shows that stereotype threat could "spill over" from math problems to other tests.

In short, words got in the way of numbers: Thoughts about the stereotype and worries about performance disrupted the subjects' ability to think through the problems. And "attempts to suppress such thoughts and to focus on the task at hand" took up significant mental resources, researchers said.

One thing worked to ameliorate the stereotype threat: drilling. Practicing math problems over and over reduced their load on working memory, making the process more instinctual and shielding them from the brain noise created by stereotypes. It's a way of overcoming the barrier, but it's additional work that groups not threatened by stereotypes don't have to do.